Difficulty-Net: Learning to Predict Difficulty for Long-Tailed Recognition

7 Sep 2022  ·  Saptarshi Sinha, Hiroki Ohashi ·

Long-tailed datasets, where head classes comprise much more training samples than tail classes, cause recognition models to get biased towards the head classes. Weighted loss is one of the most popular ways of mitigating this issue, and a recent work has suggested that class-difficulty might be a better clue than conventionally used class-frequency to decide the distribution of weights. A heuristic formulation was used in the previous work for quantifying the difficulty, but we empirically find that the optimal formulation varies depending on the characteristics of datasets. Therefore, we propose Difficulty-Net, which learns to predict the difficulty of classes using the model's performance in a meta-learning framework. To make it learn reasonable difficulty of a class within the context of other classes, we newly introduce two key concepts, namely the relative difficulty and the driver loss. The former helps Difficulty-Net take other classes into account when calculating difficulty of a class, while the latter is indispensable for guiding the learning to a meaningful direction. Extensive experiments on popular long-tailed datasets demonstrated the effectiveness of the proposed method, and it achieved state-of-the-art performance on multiple long-tailed datasets.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Long-tail Learning CIFAR-100-LT (ρ=10) Difficulty-Net Error Rate 34.78 # 12
Long-tail Learning CIFAR-100-LT (ρ=100) Difficulty-Net Error Rate 47.04 # 21
Long-tail Learning CIFAR-100-LT (ρ=50) Difficulty-Net Error Rate 43.1 # 14
Long-tail Learning ImageNet-LT Difficulty-Net (ResNet-10 w/o using RandAugment, single model Top-1 Accuracy 44.6 # 54
Long-tail Learning ImageNet-LT Difficulty-Net (ResNet-50 w/o using RandAugment, single model) Top-1 Accuracy 54.0 # 36
Long-tail Learning ImageNet-LT Difficulty-Net (ResNet-50 using RandAugment, single model) Top-1 Accuracy 57.4 # 24
Long-tail Learning Places-LT Difficulty-Net (ResNet-152) Top-1 Accuracy 41.7 # 10

Methods


No methods listed for this paper. Add relevant methods here